From one platform to four connected experiences

A B2B platform for restaurant operations — four roles, one product. I joined a stalled build, found the problem underneath the screens, and rebuilt the product from its foundation: one architecture, one design system, four modules in sequence.

Role

Solo product designer

Role

Solo product designer

Team

10, cross-functional

Team

10, cross-functional

Platform

Web · Mobile · Tablet

Platform

Web · Mobile · Tablet

Timeline

Mar. 2024 – Oct. 2025 (19 months)

4

Roles

200+

Screens

50+

Components

20

User Flows

1

Module in production

  1. Context

What is Waiter Jan

Restaurant operations run on dozens of parallel processes every day: taking orders, coordinating the kitchen, managing staff, serving guests, tracking performance. In most venues, these live in separate tools or on paper. Information gets passed by hand, service slows down, and owners never see the whole picture.

Waiter Jan set out to put all of it into one product, built around four roles: Administrator, Waiter, Chef, and Guest. Different users, one connected workflow — from the moment a guest picks a dish to the moment the owner reads the week's numbers.

A phone showing the menu on a cafe table
A phone showing the menu on a cafe table

One product for everyone around the table — from the guest’s phone to the owner’s dashboard.

  1. Starting point

What I walked into

Two modules: Administrator and Guest — were already in development. And development had effectively stopped for about two months.

The problem wasn't the quality of individual screens. It was that there was no system underneath them:

01

screens had been designed independently, with no shared information architecture

02

there were no complete user journeys — flows contradicted each other between roles

03

there was no design system and no reusable components

04

the CEO wasn't satisfied with the visual direction

05

developers kept coming back for clarifications, because the design never answered the next question

A restaurant platform isn't four apps. It's one workflow seen from four sides. Designed screen by screen, it could only get more tangled.

  1. Constraints

Decided against the clock

Everything below was decided inside these:

Hard development deadlines

The dominant constraint — every decision below was made against the clock.

Hard development deadlines

The dominant constraint — every decision below was made against the clock.

No formal user research

Structured interviews and usability sessions with restaurant staff weren’t in a startup budget.

No formal user research

Structured interviews and usability sessions with restaurant staff weren’t in a startup budget.

No product analytics

Nothing in the product measured behaviour.

No product analytics

Nothing in the product measured behaviour.

One approval gate

Only what the CEO signed off went into development.

One approval gate

Only what the CEO signed off went into development.

  1. Discovery

Why had it stopped?

I didn’t start by redrawing screens. I started by asking why the project had stalled with everything seemingly in place — requirements written, concepts drawn, functionality understood.

I audited what existed: documentation, previous design concepts, and the two modules already in development. Then I talked to people across three groups:

  • Stakeholders — business goals, priorities, what “done” meant;

  • An industry expert — how restaurant operations actually run;

  • Waiters — what a shift looks like and where time is lost.

None of this was formal research, and I won’t call it that. There was no budget for a study, so the ground under the diagnosis was industry expertise and conversations with the people doing the work.

The conclusion was uncomfortable: the project didn’t need another redesign. It needed a system — something connecting business requirements, user journeys and implementation. Nothing in the existing work did that.

  1. Decision

Repair or restart

Two options were on the table.

OPTION A

Repair

Clean up the existing files, keep development running, fix flows in place. Lower risk, no visible stoppage.

OPTION B

✅ Selected

Restart

Stop, redesign every flow and every screen from scratch, build the foundation first.

OPTION B

✅ Selected

Restart

Stop, redesign every flow and every screen from scratch, build the foundation first.

I proposed the restart, and the argument was cost rather than taste. Untangling that file was more expensive than rebuilding it: there were no reusable components worth preserving, the flows contradicted each other, and every module built on top would have multiplied the mess. Repair would have bought a little time and cost the rest of the project.

It was the riskier call — stopping in-flight development against hard deadlines. The CEO signed off.

  1. System

Foundation first

Before any screen, the question was what the four modules actually share. They aren’t four apps. They’re four views of the same order.

One state, not four copies. An order is never handed from role to role like a paper ticket. The guest places it in the app, or the waiter takes it at the table — and from that moment everyone works on the same record. The chef reads items and priority, and writes back readiness and failures. The waiter reads what the guest ordered and what’s ready. The loop closes on the Administrator: it defines the menu, the tables and the staff that everyone else consumes, and orders, sales and execution failures come back to it as analytics.

The four roles don’t sync with each other — one shared state keeps them in sync.

The design system, built alongside the first module

While designing the first module I built the layer underneath it: design tokens, components, interaction patterns. Not a UI kit assembled after the fact — components were written as the first module needed them, then reused by everything that followed.

By the end, four modules were running on 50+ components: 200+ screens and 20 flows. That count is unique component sets, not variants: one button with three styles across three sizes and four states is one component, not thirty-six. Almost every new screen was assembled from parts that already existed instead of demanding one of its own, and that is the whole difference between a design system and a folder of screens.

A component here is not a picture — it’s a set of states.

Administrator first, then Waiter, Chef, Guest. Administrator is the root of the system. The subscription is tied to it; the restaurant profile, the menu and the staff are created there. Every other module references entities this one defines. Designing the guest or the waiter experience before those entities exist means designing against a moving target. It was also the module already in development.

That’s an architectural argument, not a convenience. Starting at the root means the entities are settled before anything begins consuming them.

The other three modules only reference entities that exist in Administrator alone.

  1. Waiter Module

The hardest module

The waiter sits in the middle of everything. They talk to guests, send orders to the kitchen, manage tables, handle the bill, and use part of the admin functionality — all on a phone, standing up, under time pressure.

The table context never leaves the screen — the waiter never goes back for it.

Four decisions shaped the module

The table stays on screen. Table number and guest count are always visible. A waiter serving six tables can’t afford to check which order they’re in — a mistake there puts the wrong dish on the wrong table.

Bottom sheets instead of screens. Adding a note, choosing how to split a bill, picking a modifier — all of it happens in a sheet over the current context, never on a new screen. The waiter doesn’t lose their place in the order.

One action sends the order to the kitchen. Not a review step, then a confirm step, then a send step. The moment the order is right, it goes.

Splitting and merging a bill without leaving the order. The most argued-about moment of a meal is the one where a waiter can least afford to navigate somewhere else and come back.

What it cost. Rarely used functions moved behind an extra action. Anyone who needs them taps one more time. I took that trade deliberately: the interface is built for the ninety percent of a shift that’s repetitive and rushed, not for the ten percent that’s unusual.

  1. Identity

A mark built on a grid

The product carried an outdated logo that no longer matched where the platform was going. Once the core flows were designed, I rebuilt the brand — the mark, its construction, and the visual language the interface already implied.

I came to product design from nine years of graphic design and an art education. This is the section where that shows.

The logo is constructed on a grid, not eyeballed.

  1. Outcomes

Four modules approved, one in production

Development had been stalled for about two months. Six weeks after I joined, the first module went into development.

  • All four modules were approved by the CEO and handed over to development.

  • One module reached production, with the first venue running on it by the time I left.

  • The product is still in development today.

The team cleared the standstill together. My part was the diagnosis, the new design strategy, the architecture and the design system.

“

She built the entire UI/UX from scratch, designed a cohesive UI component library, and delivered responsive, animated, and fully prototyped designs. […] her impact on the project was outstanding.

What I can’t tell you. There are no post-launch product metrics. Analytics were never set up, so I can’t claim the design made service faster or orders more accurate. What I can show is a team that started moving again and a product that shipped.

  1. Learnings

What I'd do differently

I’d insist on usability testing with real users. I talked to stakeholders, an industry expert and waiters, but the design itself was never tested in front of the people who would use it mid-shift. There was no budget for it.

Cover of Don't Make Me Think, Revisited by Steve Krug

“Testing one user is 100 percent better than testing none.”

Steve Krug, Don’t Make Me Think

A modest budget changes how you test, not whether you test. Today I’d run at least five sessions in the real setting: waiters working through a clickable prototype in a working restaurant, not a meeting room. They cost little next to building the wrong thing — and they settle arguments with stakeholders before they start.

A modest budget changes how you test, not whether you test. Today I’d run at least five sessions in the real setting: waiters working through a clickable prototype in a working restaurant, not a meeting room. They cost little next to building the wrong thing — and they settle arguments with stakeholders before they start.

Testing with real users isn’t an extra for well-funded teams. It’s what a product stands on.

Thanks for reading

Let’s talk about your product

I design products people use at work under pressure — and I start with structure, not screens. If you’re building one, write to me. I’d rather talk about yours than mine.

I design products people use at work under pressure — and I start with structure, not screens. If you’re building one, write to me. I’d rather talk about yours than mine.

Route One case study cover: a phone with a truck navigation map on a truck dashboard

Next case

Forty tons, highway speed, six inches of screen

Route One Go · Logistics Tech · 2026

→

Product Designer

Shaping ideas into user-centric design

© 2026 Alina Sur

Fueled by coffee — designed & built by me

Product
Designer

Shaping ideas into user-centric design

© 2026 Alina Sur

Fueled by coffee — designed & built by me

Product Designer

Shaping ideas into user-centric design

© 2026 Alina Sur

Fueled by coffee — designed & built by me

Product Designer

Shaping ideas into user-centric design

© 2026 Alina Sur

Fueled by coffee — designed & built by me

Create a free website with Framer, the website builder loved by startups, designers and agencies.